Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/105523
Title: Evaluasi Keandalan Luaran Model Subseasonal to Seasonal (S2S) dalam Memprediksi Hujan Ekstrem
Other Titles: Reliability Evaluation of Subseasonal to Seasonal (S2S) Model Output in Predicting Extreme Rainfall
Authors: Faqih, Akhmad
Arif, Mochamad Fikri
Issue Date: 2021
Publisher: IPB University
Abstract: Kejadian cuaca/iklim ekstrem dapat memberikan pengaruh yang signifikan terhadap kehidupan manusia, seperti curah hujan ekstrem yang dapat mengakibatkan bencana hidrometeorologi. Karakteristik curah hujan yang berbeda-beda menyebabkan perbedaan nilai ambang batas curah hujan ekstrem pada setiap daerah. Informasi prediksi kejadian curah hujan ekstrem diperlukan untuk mengurangi dampak kerugian. Pada penelitian ini, data model prediksi S2S diuji pada wilayah yang memiliki pola hujan monsunal yaitu Kabupaten Bandung dan Kabupaten Garut. Penelitian ini bertujuan mengetahui keandalan data luaran-luaran model S2S dalam memprediksi kejadian curah hujan ekstrem. Nilai ambang batas curah hujan ekstrem ditentukan dengan metode peak over threshold. Nilai threshold (u) dihitung dengan metode persentil ke-95 dan metode persentil ke-99 untuk u-95 dan u-99, berturut-turut. Teknik koreksi bias statistik distribution mapping digunakan pada data model S2S untuk mengoreksi bias distribusi yang terdapat pada model. Berdasarkan hasil analisis metode peak over threshold diketahui bahwa u- 95 dan u- 99 dapat dijadikan acuan dalam menentukan nilai ambang batas curah hujan ekstrem dengan melihat sebaran data ekstrem mengikuti distribusi generalized pareto. Luaran model S2S memiliki akurasi yang berbeda pada setiap stasiun pengamatan dan setiap threshold curah hujan ekstrem. Prediksi luaran model S2S pada u-95 menunjukkan hasil yang lebih baik dalam memprediksi curah hujan ekstrem daripada u-99. Luaran model UKMO menunjukkan hasil evaluasi yang lebih baik dibanding model lainnya berdasarkan batasan ekstrim u-95.
Extreme weather/climate events could cause a significant impact on human life, such as extreme rainfall that can cause hydrometeorological hazards. Different precipitation characteristics cause diverse extreme rainfall threshold in every region. Information on prediction of extreme rainfall events is needed to reduce the negative impact. In this study, the S2S prediction model outputs are tested in two regions that has monsoonal rainfall pattern, i.e., Bandung Regency and Garut Regency. This study aims to evaluate the bias-corrected S2S prediction model outputs in predicting extreme rainfall events in those regions. The threshold value for extreme rainfall is obtained by using the peak over threshold method. The threshold value (u) were calculated using the 95th percentile method and the 99th percentile method for u-95 and u-99, respectively. A statistical bias correction technique, namely distribution mapping method was used to correct the distribution bias contained in the S2S model outputs. Based on the analysis of the peak over threshold method, the u-95 and u-99 could be used in determining the threshold value of extreme rainfall events that followed the generalized pareto distribution. This study shows that the S2S prediction model outputs have different accuracy for each observation station and extreme rainfall threshold. The prediction of S2S model output using u-95 threshold shows better results in predicting extreme rainfall events rather than using u-99 threshold. The output of the UKMO model shows better evaluation results than other models based on the extreme threshold u-95.
URI: http://repository.ipb.ac.id/handle/123456789/105523
Appears in Collections:UT - Geophysics and Meteorology

Files in This Item:
File Description SizeFormat 
Cover, Lembar pengesahan, Prakata, Daftar isi.pdf
  Restricted Access
Cover959.18 kBAdobe PDFView/Open
G24150063_Mochamad Fikri Arif.pdf
  Restricted Access
Fullteks1.35 MBAdobe PDFView/Open
Lampiran.pdf
  Restricted Access
Lampiran197.31 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.